Why Deep Learning can revolutionise video surveillance

Issue 4 2020 CCTV, Surveillance & Remote Monitoring

As technology takes over more aspects of our personal and professional lives, companies are increasingly turning to automation to streamline simple, manual tasks and in turn, implementing advanced technology such as artificial intelligence to offload decision-making processes. The aim, of course, being to help them save the time and cost associated with having staff carry out these activities manually.

As technology takes over more aspects of our personal and professional lives, companies are increasingly turning to automation to streamline simple, manual tasks and in turn, implementing advanced technology such as artificial intelligence to offload decision-making processes. The aim, of course, being to help them save the time and cost associated with having staff carry out these activities manually.

This scenario is particularly apt in light of the current COVID-19 crisis compelling most workers to stay in their homes. With business operations hindered and productivity jeopardised, technologies like automation and artificial intelligence obviously spring to mind as potential solutions.

In the long-term, a situation where human workers aren’t needed and machines take care of everything could be appealing. However, this is simply not an option when it comes to security. While AI can provide security staff with invaluable help, eliminating human decision-making completely could cause serious incidents – and with just the technology to blame, when accidents do happen, this may put a hard stop to the exciting growth of AI in security and the business benefits it can bring. That’s why the key to an efficient, accurate and cost effective video surveillance strategy is a combination of sophisticated technology and human interpretation.

Let’s take a look at how this synergy of artificial intelligence and human input makes for a winning security approach.

Deep Learning for enhanced accuracy

Video analytics undoubtedly revolutionised video surveillance. The shift from a situation where a staff member had to constantly monitor security footage to spot potential intruders, to one where the system itself was able to alert users of suspicious behaviour was game-changing. Before, if an intrusion was missed, resulting in a break-in or damage, the footage had to be watched retrospectively to identify the offenders, with very little scope for success. Now, however, security operators are able to act in real time, while the incident is taking place, increasing their chances of preventing crimes or containing their consequences.

However, monitoring systems, just like individuals, can make mistakes. Tasking technology with flagging potential security threats meant that not all the alerts created were real alarms and they would often call security managers’ attention when there wasn’t a real need. False alarms not only waste monitoring staff’s time, hindering operational efficiency, but also impact their ability to identify anomalous events as they cause a kind of video blindness, making them almost jaded to the excessive number of alerts raised by the system.

Kevin Waterhouse.

It’s a problem of accuracy which, luckily, can be circumvented with the help of a subset of AI – Deep Learning. While video analytics powers the detection of events, Deep Learning can play a key role in enhancing the precision of such detection. A Deep Learning filter can be used to pre-train the system to only flag the presence or movement of humans and vehicles – events, in short, that can represent a security threat. This means wildlife or environmental conditions – which are the main causes of false alarms – would not trigger an alert. This is particularly useful for perimeter protection in sterile zones, where human footfall is very limited and elements like tree branches blowing in the wind or heavy rain could simulate an alarm in a conventional system and alert staff of a potential intruder – causing them to interrupt more important activities. When the alert is generated, that’s where human intervention is incredibly valuable. The AI-powered video analytics may spot an intruder, but it’s still the worker’s responsibility to review the situation brought to their attention by the system and determine how to respond to it.

As it often happens with technology, the hype around artificial intelligence is causing organisations across all fields to consider its implementation, fearing getting left behind and wanting to innovate the way they do business. It’s important, however, to evaluate how such technologies can improve their operations, before rushing to invest in them. When it comes to security and specifically, video surveillance, Deep Learning can provide unprecedented accuracy, minimising false alarms and ultimately, helping security managers and their teams protect premises like never before.

Share this article:
Share via emailShare via LinkedInPrint this page

Further reading:

Retail solutions beyond security
Issue 8 2020, Axis Communications SA, Technews Publishing, Hikvision South Africa , Editor's Choice, CCTV, Surveillance & Remote Monitoring
The need for security technology to deliver more than videos of people falling or stealing from retail stores is greater than ever.

A ‘step-change’ in tracking and monitoring
Issue 8 2020, Dahua Technology South Africa , CCTV, Surveillance & Remote Monitoring, Products
View an overall scene while monitoring a specific person or vehicle, or obtain a panoramic view while being able to zoom in and track particular targets.

AI-powered hardhat detection
Issue 8 2020, Hikvision South Africa , Industrial (Industry), CCTV, Surveillance & Remote Monitoring
Hardhats save lives, but only if people wear them. Intelligent, AI-powered hardhat cameras are helping to ensure workers in dangerous locations stay safe at all times.

LPR is an integrated solution
CCTV Handbook 2021, DeepAlert, XtraVision, Technews Publishing , CCTV, Surveillance & Remote Monitoring
The power of LPR is only experienced when it is integrated with other technologies, from video analytics and AI, through to connecting to external databases.

Video storage focused on Milestone
CCTV Handbook 2021 , CCTV, Surveillance & Remote Monitoring
Video Storage Solutions (VSS) was formed with the single objective of providing system integrators who are part of the Milestone Community with verified video recording and storage appliances through authorised distributors.

Thermal perimeter and fire protection
CCTV Handbook 2021, Hikvision South Africa , CCTV, Surveillance & Remote Monitoring
Hikvision has launched HeatPro, a new thermal series of cameras designed to provide affordable all-weather deterrent and alerts, specifically aimed at perimeter protection and fire prevention applications.

360° thermal detection
CCTV Handbook 2021, XtraVision , CCTV, Surveillance & Remote Monitoring
Thermal Radar is a network-based thermal surveillance and detection solution providing constant 360° situational awareness.

Cloud surveillance options from AWS
CCTV Handbook 2021, Technews Publishing , CCTV, Surveillance & Remote Monitoring
The idea of doing away with the task of managing your own IT infrastructure and simply using a reliable and scalable infrastructure, provided by and managed by someone else, seems like a good idea.

Video security primer released
CCTV Handbook 2021, Eagle Eye Networks , CCTV, Surveillance & Remote Monitoring
The right video management system can make an organisation smarter and more efficient, as well as more secure according to Eagle Eye Networks’ new guide in evaluating cloud video surveillance systems.

Flexibility for parking management
CCTV Handbook 2021 , CCTV, Surveillance & Remote Monitoring
With intelligent video analytics smart security cameras can provide parking operators with increased flexibility using an open IoT platform that allows security cameras to run LPR applications as well as vehicle counting, fire/smoke and hazard detection and integrations with payment solutions.